四旋翼飞行器姿态传感器的故障诊断技术研究

发布时间：2018-03-28 08:46

本文选题：陀螺仪　切入点：BP神经网络　出处：《中北大学》2017年硕士论文

【摘要】：在四旋翼无人机控制系统中,姿态传感器起到了测量相关参数的作用,然后将参数反馈到控制系统中进行姿态解算,实现无人机的姿态控制。姿态传感器作为飞控系统的重要组成部分,它的状态将关系到无人机的稳定飞行,所以对姿态传感器的故障诊断是非常有意义的。本文主要研究工作如下:本文通过比较现有各种故障诊断技术的优势与缺陷,并考虑到四旋翼飞行器飞控系统的特性,将BP神经网络诊断方法应用于四旋翼飞行器飞控系统并针对BP神经网络的收敛速度慢、精度低等缺点,采用遗传算法进行改进。将陀螺仪三个角度输出作为估计参数,建立了翻滚角观测器、俯仰角观测器和偏航角观测器,在线估计陀螺仪输出。对于故障类型辨别的问题,本文将故障数据与传感器实测数据进行一元线性回归分析,能够成功辨别恒增益与恒偏差故障。并提出了双阈值检测方法,以提高故障检测的准确率。使用的不同训练算法与隐层节点对BP神经网络进行训练,以选取最优训练算法与隐层节点数;同时对GA-BP网络性能进行仿真,数据表明GA-BP网络在收敛速率和精度方面有了明显提升。在MATLAB仿真平台下,进行故障检测和辨别仿真,结果表明,本文所使用的方法不仅能够准确的预测陀螺仪的输出,而且双阈值检测故障法也能较为准确的检测到故障注入点,并对其进行标记,并且一元线性回归法也能准确的辨别故障类型。
[Abstract]:In the four-rotor UAV control system, the attitude sensor acts as a measure of related parameters, and then the parameters are fed back to the control system for attitude calculation. As an important part of flight control system, attitude sensor will affect the steady flight of UAV. So it is very meaningful to fault diagnosis of attitude sensor. The main research work of this paper is as follows: this paper compares the advantages and disadvantages of various fault diagnosis technologies, and takes into account the characteristics of flight control system of four-rotor aircraft. The BP neural network diagnosis method is applied to the flight control system of four-rotor aircraft. The genetic algorithm is used to improve the BP neural network because of its slow convergence speed and low precision. The output of three angles of the gyroscope is taken as the estimation parameter. The roll angle observer, pitch angle observer and yaw angle observer are established to estimate the gyroscope output online. The method of double threshold detection is proposed to improve the accuracy of fault detection. Different training algorithms and hidden layer nodes are used to train BP neural network. In order to select the optimal training algorithm and the number of hidden layer nodes, at the same time, the performance of GA-BP network is simulated. The data show that the convergence rate and precision of GA-BP network have been improved obviously. Under the MATLAB simulation platform, fault detection and discrimination simulation are carried out. The results show that the method used in this paper can not only accurately predict the output of gyroscope, but also detect the fault injection point accurately and mark it. And the linear regression method can also accurately identify the type of fault.
【学位授予单位】：中北大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：V267;TP183

【参考文献】